Figure 3
Model performance comparison among models trained on different silver-positives. Self-potential (green bars) and self-BoPD (blue bars): silver-positives were built via applying the LR model trained on gold-label data on potential BoPD cohort and EHR diagnosed BoPD cohort, respectively, to generate predicted positives. Silver (red bars): rule-based selection of silver-positives from EHR diagnosed BoPD cohort, also called knowledge-enriched silver-positives. Black verticle lines represent 95% confidence interval.
